1 paper · 1 filter
Jinman Zhao, Xueyan Zhang, Xingyu Yue +3
Current common interactions with language models is through full inference. This approach may not necessarily align with the model's internal knowledge. Studies show discrepancies…